Integrated Communication and Computation-Enabled Cooperative Spectrum Sensing with the Mamba Network

Cooperative spectrum sensing (CSS), which serves as a foundation mechanism, enables cognitive radio (CR) to mitigate inter-system interference. However, existing CSS schemes overlook the inevitable noise in the reporting channel and neglect spectral resource consumption associated with reporting. Furthermore, existing CSS schemes that adopt Long Short-Term Memory (LSTM) for recognizing the primary user (PU) activity pattern face limitations in parallel training, resulting in prolonged training time. To address these challenges, an integrated communication and computation (ICC)-enabled scheme is proposed in this paper, termed as ICC-Mamba, which utilizes a convolutional neural network (CNN) to extract spatial features and integrates a Mamba network to discern PU activity patterns, i.e., temporal features. Extensive simulation results demonstrate that the ICC-Mamba scheme outperforms existing CSS schemes in terms of detection performance, resource consumption efficiency, and robustness to signal-to-noise ratio (SNR) in both sensing and reporting channels.